The six categories, side by side
Every real SEO deliverable belongs to one of six categories. Agencies bundle them, rename them and reorder them in decks, but the underlying work doesn't change.
The taxonomy matters for one practical reason: it lets you read a proposal and see what's missing. A quote with no link acquisition is a content service. A quote that's 80% technical for a 30-page site is an upsell.
| Category | The work | What it changes | Share of a typical retainer |
|---|---|---|---|
| Technical SEO | Crawling, indexing, site speed, Core Web Vitals, schema, redirects, JavaScript rendering. | Whether Google can reach, render and trust your pages at all. | 20–30% up front, ~10% after |
| On-page SEO | Titles, headings, internal links, URL structure, per-page schema, image handling. | How well a page you already have matches a query you already want. | 10–15% |
| Content | Keyword research, briefs, new pages, rewrites of pages that have decayed. | How many queries you're eligible to answer in the first place. | 35–45% |
| Links and digital PR | Earned placements, data stories, expert commentary, partnerships, resource pages. | Whether you can compete for the queries everyone else also wants. | 20–30% |
| Local SEO | Google Business Profile, citations, reviews, location pages, map pack visibility. | Whether you show up when somebody nearby searches with intent. | Often the entire budget for a single-location business |
| International SEO | Hreflang, country targeting, subfolder or ccTLD structure, translated and localised copy. | Which country's version of your page ranks in which market. | Zero unless you sell across borders |
Technical SEO: the ceiling, not the engine
Technical SEO is the only category that doesn't produce growth by itself. It removes the thing that was capping growth. That distinction changes how you should buy it.
On an older site with a migration or two in its history, the first technical pass often recovers traffic that was quietly lost years ago — no new content required. On a clean 40-page site built last year, the same pass finds four things and takes an afternoon. Same service, wildly different value.
- Crawl and index control — robots directives, sitemaps, canonicals, pagination, index bloat.
- Rendering — whether your content exists before JavaScript runs. This is the single most common silent failure on modern stacks.
- Speed and Core Web Vitals — LCP, INP, CLS. Real thresholds, measurable, and mostly a developer's job rather than a marketer's.
- Redirects and site structure — chains, loops, orphan pages, and whatever survived the last replatform.
- Schema markup — how you tell search engines what a page is, rather than hoping they infer it.
On-page and content: the two people keep merging
These get bundled constantly, and they're genuinely different purchases with different economics.
On-page is per-page and finite
On-page work takes a page that already exists and makes it a better answer to a query it's already close to. Title, headings, internal links, schema, the first hundred words.
- You can buy it as a one-off project — 200 pages audited and rewritten, done, invoice closed.
- It's the fastest-moving category. Search Console impressions can shift inside a few weeks.
- It runs out. Once every page is properly built, there's nothing left to fix until you publish more.
Content is per-query and never finishes
Content creates eligibility. You cannot rank for a query you haven't written about, and no amount of on-page work on your homepage changes that.
It's also the only category where cadence beats volume. Four solid pages a month for a year outperforms fifty in one quarter followed by silence, because search engines reward sites that keep being useful and because internal linking compounds.
Why the difference matters when you're buying
Old site with traffic, untouched for three years? Buy on-page first — those pages are already indexed and already carry some authority. New or thin site? On-page has nothing to act on. Buy content, then buy on-page for that content six months later.
Links and digital PR: slowest, priciest, hardest to fake
This is where the gap between honest and dishonest agencies is widest, because the output is the hardest to inspect.
Real link work is relationship work at volume: pitching journalists, publishing something worth citing, commenting as a named expert. It's slow, it doesn't scale with budget, and a good month on a ₹75,000 retainer is two to five earned placements. Anyone promising forty is selling something else.
- Ask where the links come from. "Our network" means a private blog network. That's a penalty waiting to land on your domain, not theirs.
- Ask for URLs, not counts. A monthly report listing 12 links without a single clickable URL is a report about nothing.
- Watch the referring-domain graph. Slow and steady is healthy. A sudden spike of hundreds is the signature of bought links.
- Expect it to be expensive. Links are the category where costs are highest per unit and least negotiable — see what link building actually costs.
Local and international: the two categories shaped by geography
These two aren't add-ons to the other four. They're separate disciplines with their own tooling, their own ranking surfaces and their own failure modes.
Local SEO
Local SEO competes for the map pack — the three-result block that appears above the normal results for anything with local intent. It's a different ranking surface with different inputs, which is why local SEO can work brilliantly for a business whose website barely ranks at all.
- Google Business Profile: categories, services, hours, photos, posts, Q&A.
- Reviews — volume, recency and whether anyone replies.
- NAP consistency across directories, which matters more in India than most people expect because listing data is so messy here.
- Location pages with genuinely different content, not one template with the city name swapped.
International SEO
Almost entirely structural. Get hreflang, URL structure and country targeting right and translated content works. Get them wrong and your Indian pages outrank your UAE pages in Dubai, permanently, however good the copy is. Don't buy this until you have real demand in a second market — the structure is expensive to build and painful to unwind.
Which two to buy first, by where your site actually is
Nobody buys all six at once, and the sequence matters more than the total budget. This is the table worth arguing with your agency about.
- Two rules sit under that whole table. Never buy content before the site can be crawled and rendered.
- And never buy links before there's something on the site a stranger would willingly cite.
| Where your site is | Buy first | Buy second | Skip for now |
|---|---|---|---|
| Brand new, under 20 pages, no rankings | A small technical foundation, once | Content on the ten queries closest to money | Links, digital PR, international |
| 1–3 years old, some rankings, growth flat | On-page and content refresh | Internal linking across what you already have | Net-new content volume |
| Old, thousands of pages, traffic falling | Technical audit and index control | Consolidating pages that compete with each other | New content until the crawl is clean |
| Single-location service business | Local SEO — profile, reviews, location pages | A handful of properly built service pages | Digital PR and national keyword sets |
| Selling into two or more countries | Hreflang and country structure | Market-specific content, not translations | Anything at all before the structure is right |
What gets sold as a category but isn't one
Four things show up in proposals as line items and shouldn't. If you see them, ask which of the six they map to.
- "SEO optimisation" — a word standing in for a scope. Ask for the line items and the volumes. The full list is in what SEO services include.
- Directory submissions at volume — 200 listings on sites nobody visits. This was a tactic in 2011.
- Social signals — likes and shares are not a ranking input you can buy your way into. Social is a real channel; it just isn't this one.
- Reporting as a category — reporting wraps the other six. It's a deliverable, not a purchase, and it shouldn't be more than about a tenth of the hours.